I have estimated two-country SOE in a currency union a la Gali/Monacelli with habit formation and price indexation. However, since the model is highly stylized, measurement errors tend to be large. Is there a way in Dynare to calibrate the variance of the measurement errors to reasonable values, so that, the fundamental shocks explain, say, 70%-80% of the variation in the data.
Usually you use priors for that. Use e.g. a uniform prior for the stderr of the measurement error and set the upper bound to the desired value, i.e. fraction of the data variance (which you have to compute yourself).
You cannot do that automatically.
What you can do is to estimate the measurement error using a tight prior distribution (this is almost equivalent to calibrating), see if the results are in line with what you want and, if not, refine and retry.